Validating the Literature

Validating the Literature

A verification of COVID-19's long-term effects through anomaly detection

Validating the Literature

A verification of COVID-19's long-term effects through anomaly detection

Devin Lim, Ponnu Soman, Duong Le, Lakshit Gupta

Background

Background

COVID reported death-toll: ~1,200,000^[1]

Background

COVID reported death-toll: ~1,200,000^[1]

Background

COVID reported death-toll: ~1,200,000^[1]

→ But wait, there's more!

What the Literature Says

What the Literature Says

• Measurable effects long after the initial infection
• Damages every part of the body
  • Brain
  • Lungs
  • Heart
  • Kidneys
  • And more!

→ How?

Mechanism

Mechanism

1. Damage to the mitochondria

Mechanism

1. Damage to the mitochondria → nectroptosis

Mechanism

1. Damage to the mitochondria → nectroptosis → organ damage^[2]

Mechanism

1. Damage to the mitochondria → nectroptosis → organ damage^[2]
2. Excessive immune response

Mechanism

1. Damage to the mitochondria → nectroptosis → organ damage^[2]
2. Excessive immune response → inflammation & blood clots

Mechanism

1. Damage to the mitochondria → nectroptosis → organ damage^[2]
2. Excessive immune response → inflammation & blood clots → more organ damage^[3]

Mechanism

1. Damage to the mitochondria → nectroptosis → organ damage^[2]
2. Excessive immune response → inflammation & blood clots → more organ damage^[3]

→ Seems... worrying

The Big Question:

The Big Question:

Are the claims true?

Methodology

Methodology

A little revisit:

Methodology

A little revisit:

Is this accurate?

How is it counted anyway?

Death Certificates

Death Certificates

Any mention of COVID == 1 count

Problems?

Problems?

• Reporting Lag
• Human Error
• Indirect, delayed deaths
• Inconsistency between jurisdictions
• Limited testing capacity, especially early in the pandemic
• And so on...

→ Underreporting!

Solution?

Solution?

Excess Mortality

Excess Mortality

Idea: Death rate is fairly constant year-over-year

Excess Mortality

Idea: Death rate is fairly constant year-over-year

Excess Mortality

Steps:

1. Ignore reported count
2. Calculate expected deaths
3. Compare with actual deaths

Result:

1. A sudden, drastic increase in deaths
2. Pattern follows reported deaths
3. An undercount of ~200,000!

→ Can we do the same with other diseases?

Data Methodology

Data Methodology

1. Data Sourcing

2. Data Cleaning

3. Prediction Methods

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Chronic Health Indicators

Behavioral Risk Factor Surveillance System

• Respiratory diseases
• Cardiovascular diseases
• Depression
• Diabetes
• Kidney
• Cancer

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Chronic Health Indicators

Behavioral Risk Factor Surveillance System

• Respiratory diseases
• Cardiovascular diseases
• Depression
• Diabetes
• Kidney
• Cancer

Prevalence of Disability Status and Types

Disability and Health Data System

• Overall disability rates
• Breakdown by disability types

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Chronic Health Indicators

Behavioral Risk Factor Surveillance System

• Respiratory diseases
• Cardiovascular diseases
• Depression
• Diabetes
• Kidney
• Cancer

Prevalence of Disability Status and Types

Disability and Health Data System

• Overall disability rates
• Breakdown by disability types

Daily COVID-19 Vaccines Administered

Our World in Data + CDC

• Vaccines

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Chronic Health Indicators

Behavioral Risk Factor Surveillance System

• Respiratory diseases
• Cardiovascular diseases
• Depression
• Diabetes
• Kidney
• Cancer

Prevalence of Disability Status and Types

Disability and Health Data System

• Overall disability rates
• Breakdown by disability types

Daily COVID-19 Vaccines Administered

Our World in Data + CDC

• Vaccines

Deaths Involving Coronavirus Disease

Department of Health & Human Services

• COVID death-toll

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Standard stuff: filtering, renaming, grouping, sorting

brfss_filters = {
    "Locationabbr": ["UW"],
    "Topic": [
        "Asthma",
        "COPD",
        ...
    ],
    "Response": ["Yes"],
}

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Standard stuff: filtering, renaming, grouping, sorting

brfss_filters = {
    "Locationabbr": ["UW"],
    "Topic": [
        "Asthma",
        "COPD",
        ...
    ],
    "Response": ["Yes"],
}
brfss_col_renames = {
    "Locationabbr": "LocationAbbr",
    "Data_value": "DataValue",
    ...
}

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Standard stuff: filtering, renaming, grouping, sorting

brfss_filters = {
    "Locationabbr": ["UW"],
    "Topic": [
        "Asthma",
        "COPD",
        ...
    ],
    "Response": ["Yes"],
}
brfss_col_renames = {
    "Locationabbr": "LocationAbbr",
    "Data_value": "DataValue",
    ...
}
brfss_val_renames = {
    "Question": {
        "Adults who have been told they currently have asthma (variable calculated from one or more BRFSS questions)": "Current Asthma",
        "Ever told you had angina or coronary heart disease?": "Coronary Heart Disease",
        ...
    },
    "Topic": {
        "Asthma": "Respiratory Diseases",
        "COPD": "Respiratory Diseases",
        "Other Cancer": "Cancer",
    },
}

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Spline

Data Methodology

Data Sourcing

Data Cleaning

Prediction Methods

Spline

Time Series

Chronic Health Indicators

JACI: There is evidence that COVID-19 increases the risk of developing asthma

We were not able to find evidence of a causal link between COVID-19 and COPD

American Heart Association: There is evidence that COVID-19 causes lasting heart damage

Yale: There is strong evidence of significant increases in brain-related symptoms after a COVID-19 infection. Depression is one such symptom.

Cedars-Sinai Medical Center: There is strong evidence that COVID-19 increases the risk of new-onset (type 2) diabetes

Nature: There is strong evidence that COVID-19 increases the risk of CKD

Yale: Confirmed the evidence, but also found that CKD that developed due to COVID-19 recovered faster

We found mixed literature on the causal link between COVID-19 and Cancer

Disability

Vaccine

Conclusion

• Vaccine works!
• All time highs for:
  • Disability
  • CKD
  • Diabetes
  • Depression
  • Asthma
  • Non-skin cancers
• Agreed with the literature on:
  • CKD
  • Disability
  • Diabetes
  • Depression
  • Asthma